Article in International Journal of Entrepreneurship · December 2020 doi: 10. 4018/ijegcc. 2020070105 citations 11 reads 286 João Manuel Pereira Polytechnic Institute of Setúbal Instituto Politécnico de Lisboa 95

Methods

Theresearchwasconductedoverseveralstages.Thesamplehasbeencharacterizedthroughdescriptive statistics,fromtheadhesiontestofJarqueandBera(1980),togetherwithresidualstabilityplots.Since theestimationsinvolvetimeseries,itisnecessarytoexaminethestationarynatureofthedataseries usingtheLevin,Lin,andChu(2002),ADF(DickeyandFuller,1981),PP(PierrePerronandPhillips,

1988)tests.Toestimatethefinancialmarkets’breakdowns,theauthorsusedtheunitroottestby Clementeetal.(1998).ToassessthesignificanceofthecausalrelationshipsbetweentheCOVID-19 dataseriesandthefinancialmarketsunderanalysis,theauthorsusedtheVARGrangerCausality/ BlockExogeneityWaldTestsmodel,applyingWald’sstatisticstotestwhetherthenullhypothesis thatthecoefficientsofthelaggedendogenousvariablesofthevariable“cause”arenullorarenot “causing,”inGrangerianterms,thedependentvariable.However,itisimportanttoemphasizethat thistest’sresultishighlysensitivetothenumberoflagsconsideredinthemodel.Consequently,the firstconcernistoproperlyestimatethisvaluetoarriveatrobustevidence(Gujarati,2004).Also,and todeterminethenumberoflagstoincludeinthecausalitytests,theauthorsusedtheHQ(HannanQuinninformationcriterion)criterion.Toassesstherobustnessofthemodel,weappliedtheVAR ResidualSerialCorrelationLMTests.AlthoughHurst’sexponentisnotuseddirectly,weapplied amethodologythatindirectlyproposesthesameinformation,namelytheDetrendedFluctuation Analysis(DFA).DFAisananalysismethodthatexaminestimedependenceonnon-stationarydata series.Thistechnique,assumingthatthetimeseriesarenon-stationary,avoidsspuriousresultswhen theanalysisfocusesonthedataseries’srelationshipsinthelongrun.Thismethodologywasdeveloped byPengetal.(1994),havingoriginatedinthestudyofDNAbehavior.Thismethodwaslaterused toexaminethebehavioroffinancialseries.TheDFAhasthefollowinginterpretation:0<α<0.5: anti-persistentseries;α=0.5serieshasarandomwalk;0.5<α<1persistentseries.

Thepurposeofthistechniqueistoexaminetherelationshipbetween x_k and x_{k t+} valuesat differenttimes(Ferreira,Dionísio,Guedes,andZebende,2018).

Adatasetx_k ,withk = 1,...,t ,shouldbeconsideredequidistantobservations.DFA’sfirststep istobuildanewseries:

x t( ) = ∑^tk=1x_k  (1)

Thesecondstepinvolvesobtainingthe z t( )trendofeachfractionbyusingtheleast-squares methodandobtainingtheseriessubtractedfromthetrend(detrended),thatis: